High-temperature performance of metakaolin enhanced recycled aggregate concrete: mechanical, microstructural, and machine learning prediction

Abstract This study examines the thermomechanical behavior of concrete incorporating recycled fine and coarse aggregates 0–100% replacement derived from parent concrete containing 5% metakaolin. Specimens were subjected to temperatures between 20 and 800 °C, and their residual strength, fracture energy, mass loss, and microstructural evolution were evaluated. Results show that moderate recycled aggregate (RA) contents 25–50% maintain performance comparable to natural aggregate concrete due to metakaolin-induced matrix densification and ITZ refinement, while higher RA levels 75–100% exhibit pronounced deterioration with losses exceeding 80–95% at higher temperatures depending on recycled aggregate (RA) content, linked to porous adhered mortar and weakened interfacial zones. Fracture energy improved up to 400–500 °C across all mixes before declining sharply beyond 600 °C, reflecting the onset of CSH decomposition and microstructural destabilization. Mass loss trends further confirmed reduced thermal stability in RA rich mixtures, reaching 12–13% at 800 °C compared to < 6% in the control. Response surface modeling and machine learning regression revealed nonlinear degradation thresholds and demonstrated strong predictive capability for thermomechanical performance. SEM analysis revealed a substantial increase in porosity/crack density, confirming progressive microstructural degradation. Machine learning models demonstrated high predictive accuracy, with the XGBoost model achieving an R 2 value of 0.998 and RMSE of 0.79, outperforming other models. The findings highlight that metakaolin-modified recycled aggregates, when used at moderate replacement levels, can produce thermally resilient and sustainable concretes suitable for fire exposed or high-temperature applications.